DCVLM-Baseline is the reference training mixture from our DataComp-VLM paper. It is a pre-mixed, decontaminated, ready-to-train multimodal pretraining dataset, materialized as flat WebDataset tar shards so it can be consumed by any training stack. This is a 200B-token dataset release consisting of 103,985,276 samples, curated from our DCVLM-large data pool. A smaller 6.25B-token version is also available. The mixture matches our paper's Instruction-heavy mix. Proportions are by training samples: Each training example is one WebDataset sample: a group of members sharing a common key. Members: Text-only samples (datatype == text) carry no image members. Multi-image samples (mmdocmt, and some…
SAVRN Model Hub · Datasets by Task
Image and text to text Datasets
3 datasets in the SAVRN Model Hub for image and text to text, from publishers including ML Foundations, Anonymous, HuggingFaceM4.
3 datasets.
FineVision is a massive collection of datasets with 17.3M images, 24.3M samples, 88.9M turns, and 9.5B answer tokens, designed for training state-of-the-art open Vision-Language-Models. More detail can be found in the blog post: https://huggingface.co/spaces/HuggingFaceM4/FineVision The version in this repository concatenated all the configs in the original dataset and then shuffled them. This is done to facilitate streaming the data directly from the hub! Each of the publicly available sub-datasets present in FineVision are governed by specific licensing conditions. Therefore, when making use of them you must take into consideration each of the licenses governing each dataset. To the…
RL prompt corpus of the ICLR 2027 submission MemGUI-RL (project page: https://memgui-rl-anonymous.github.io/). Every record is one annotated state of a MemGUI-3K trajectory in the ConAct conversation format consumed by the FARPO trainer (https://github.com/memgui-rl-anonymous/MemGUI-RL): Each record has taskid, stepnumber, ispositive, badstep, rawresponse (the annotated ConAct response:,,,, ), conversations (system prompt, user prompt with the folded history / UI memory / recent step record and the screenshot, reference assistant turn) and metadata (impact / reasonableness labels of the step, trajectory ids). Screenshots are not duplicated here: image references have the form…